To identify baseline characteristics predictive of visual acuity (VA) outcomes at month 12 (M12) and treatment frequency in the first 12 months of the phase III HARBOR study.
Retrospective, exploratory analysis of multicenter randomized controlled trial data.
setting : Randomized, multicenter. study population : Patients aged ≥50 years with subfoveal wet age-related macular degeneration (AMD) who had best-corrected VA (BCVA) values measured at baseline and M12. intervention : Intravitreal ranibizumab 0.5 mg administered monthly (n = 249) or as needed (PRN) after 3 monthly loading doses (n = 251). main outcome measures : BCVA change from baseline at M12, percentage of patients who gained ≥15 letters (3 lines) in BCVA from baseline at M12, and percentage of patients who achieved ≥20/40 vision (Snellen) at M12 served as the basis for analyzing baseline predictors of observed VA outcomes in the monthly and PRN groups. Total number of ranibizumab PRN injections in the first 12 months was also evaluated. Only variables that were statistically significant ( P < .05) remained in the final statistical models.
Baseline predictors of BCVA change from baseline at M12 and/or percentage of 3-line gainers included lower BCVA, younger age, smaller total choroidal neovascularization (CNV) leakage area, smaller area of occult CNV, and presence of subretinal fluid (SRF). Baseline predictors of ≥20/40 vision at M12 included higher BCVA, smaller total CNV leakage area, and presence of SRF. SRF thickness >118.25 μm at baseline predicted requiring more ranibizumab injections in the first 12 months of treatment.
Select baseline characteristics have predictive value for visual prognosis and treatment frequency in ranibizumab-treated patients with wet AMD.
Although most patients with neovascular age-related macular degeneration (wet AMD) experience improvements in visual and anatomic outcomes with anti–vascular endothelial growth factor (anti-VEGF) therapy, a subset of patients do not respond optimally to treatment. For example, despite significant visual improvements in the majority of patients treated with ranibizumab (Lucentis; Genentech, Inc, South San Francisco, California, USA) in the ANCHOR (Anti-VEGF Antibody for the Treatment of Predominantly Classic Choroidal Neovascularization in Age-Related Macular Degeneration) and MARINA (Minimally Classic/Occult Trial of the Anti-VEGF Antibody Ranibizumab in the Treatment of Neovascular AMD) trials, ∼10% of patients in ANCHOR and ∼8%–10% of patients in MARINA lost ≥15 letters in best-corrected visual acuity (BCVA) from baseline at month 24. Findings from several retrospective studies suggest that certain demographic and choroidal neovascularization (CNV) lesion characteristics at baseline may increase the likelihood of having better or worse treatment outcomes. The elucidation of demographic, anatomic/imaging, and functional characteristics at baseline that can be used as predictive biomarkers of response to anti-VEGF treatments will be helpful for the successful development of next-generation treatments for wet AMD.
Historically, smaller CNV lesion sizes at baseline are generally associated with better visual acuity (VA) outcomes at month 24 in patients with subfoveal wet AMD, as shown in the TAP (Treatment of Age-Related Macular Degeneration With Photodynamic Therapy) and VIP (Verteporfin in Photodynamic Therapy) trials. Retrospective analyses of 12-month data from ANCHOR and 24-month data from MARINA demonstrated that lower baseline BCVA, smaller baseline CNV lesion size, and younger baseline age were predictive of better VA outcomes in ranibizumab-treated patients with wet AMD. These baseline predictors of visual outcomes were also observed in a retrospective analysis of ranibizumab- or bevacizumab-treated patients with wet AMD who completed the first year of CATT (Comparison of Age-Related Macular Degeneration Treatments Trials).
The HARBOR (Phase III, Double-Masked, Multicenter, Randomized, Active treatment-controlled Study of the Efficacy and Safety of 0.5 mg and 2.0 mg Ranibizumab Administered Monthly or on an As-needed Basis [PRN] in Patients With Subfoveal Neovascular Age-Related Macular Degeneration) trial provides a large prospective data set on ranibizumab-treated patients with wet AMD. At the end of the first year of HARBOR (the primary endpoint), clinically meaningful and similar improvements in BCVA were observed across all treatment groups. The mean change from baseline in BCVA at month 12, based on the last-observation-carried-forward method, was +10.1 letters (0.5 mg monthly; n = 275), +8.2 letters (0.5 mg PRN; n = 275), +9.2 letters (2.0 mg monthly; n = 274), and +8.6 letters (2.0 mg PRN; n= 273). These improvements were achieved with a mean of 7.7 and 6.9 injections in the 0.5 mg and 2.0 mg PRN groups, respectively, and a mean of 11.3 and 11.2 injections in the 0.5 mg and 2.0 mg monthly groups, respectively. Although the prespecified superiority and noninferiority comparisons were not met at year 1, the HARBOR study 1-year results demonstrated that ranibizumab 0.5 mg administered on a monthly or PRN regimen, guided by VA and strict spectral-domain optical coherence tomography (SD-OCT) re-treatment criteria, resulted in clinically meaningful VA gains in patients with wet AMD.
This retrospective, exploratory analysis of data from the HARBOR study investigated demographic and baseline characteristics predictive of VA outcomes at month 12 in the ranibizumab 0.5 mg monthly and 0.5 mg PRN groups, and treatment frequency in the first 12 months in the ranibizumab 0.5 mg PRN group. The ranibizumab 0.5 mg dose was chosen because it is the approved dosage for the treatment of wet AMD, and is thus the most useful to clinicians in the context of evaluating baseline predictors of treatment response. Understanding which demographic and ocular characteristics at baseline can be used as predictive biomarkers of response to anti-VEGF treatment may help retina specialists design optimal treatment plans tailored from the start of therapy to individual patients with wet AMD.
Study Design and Patient Eligibility
The methods for the HARBOR study have been reported previously. Briefly, HARBOR was a 24-month, phase III, randomized, multicenter, double-masked, active-treatment controlled study that evaluated the efficacy and safety of intravitreal ranibizumab 0.5 mg and 2.0 mg administered monthly or PRN after 3 monthly loading doses in treatment-naïve patients (n = 1097) aged ≥50 years with active subfoveal wet AMD (predominantly classic, minimally classic, or purely occult CNV) and baseline BCVA of 20/40–20/230 (Snellen equivalent). Re-treatment in the PRN groups was based on BCVA using Early Treatment Diabetic Retinopathy (ETDRS) charts and SD OCT criteria. The study was conducted in accordance with Good Clinical Practice guidelines (International Conference on Harmonization of Technical Requirements for Registration of Pharmaceuticals for Human Use E6), applicable US Food and Drug Administration regulations, and the Health Insurance Portability and Accountability Act. The study protocol was approved by institutional review boards at each location prior to study initiation, and all participants provided written informed consent. The HARBOR study is registered at ClinicalTrials.gov (no. NCT00891735 ).
The primary endpoint of the HARBOR study was the mean change from baseline in BCVA at month 12. In this retrospective, exploratory analysis of multicenter, randomized, controlled data from the HARBOR trial, the main outcome measures that served as a basis for baseline predictors of VA outcomes at month 12 were BCVA change from baseline at month 12, the proportion of patients with a BCVA gain of ≥15 ETDRS letters from baseline at month 12, and the proportion of patients with a Snellen equivalent of 20/40 or better at month 12 in the ranibizumab 0.5 mg monthly and 0.5 mg PRN groups. A post hoc analysis of the total number of injections in the first 12 months in patients receiving ranibizumab 0.5 mg PRN was also performed.
These exploratory analyses were performed on observed data from patients in the ranibizumab 0.5 mg monthly (n = 249) and 0.5 mg PRN (n = 251) treatment groups who had BCVA values measured at both baseline and month 12. A multivariate linear regression model was performed to determine baseline predictors for BCVA change from baseline at month 12. Multivariate analysis using logistic regression models was performed to determine baseline predictors for BCVA gain of ≥15 letters from baseline at month 12 and baseline predictors for Snellen equivalent of 20/40 or better at month 12. Proportional odds and linear regression models were used to determine baseline predictors for the total number of injections in the 0.5 mg PRN group during the first 12 months of the study. Final models included only significant ( P < .05) predictors selected by stepwise selection methods. All analyses were exploratory in nature and no multiplicity adjustment was made. All analyses were conducted using SAS software version 9.2 (SAS Institute, Inc, Cary, North Carolina, USA).
To better facilitate interpretation of the data and illustrate the trends graphically, a Classification and Regression Tree (CART) analysis was used to dichotomize values for continuous baseline predictors from the final models where appropriate. All cutoff values were derived using CART v6.0 (Salford Systems, San Diego, California, USA). A total of 44 variables (demographic parameters, baseline characteristics, and assessments) were included in the statistical modeling; for predictors of visual outcome, regimen (monthly vs PRN) was also included in the candidate list.
Patient demographics and baseline ocular characteristics for the study eye, previously reported, were well balanced among the ranibizumab treatment arms. In the 0.5 mg monthly and 0.5 mg PRN groups, mean VA at baseline was 54.2 and 54.5 letters (approximate Snellen equivalent of 20/80), respectively, and the mean age at baseline was approximately 79 years. Overall, approximately 46% of patients in the 0.5 mg monthly and 0.5 mg PRN groups had minimally classic CNV lesions, 16% had predominantly classic lesions, and 37% had purely occult CNV. Total area of CNV leakage and total area of CNV in the 2 treatment groups ranged from 3.31 to 3.48 disc area (DA) and from 3.04 to 3.27 DA, respectively. The mean baseline area of occult CNV in the 0.5 mg monthly and 0.5 mg PRN arms were 2.81 DA and 2.59 DA, respectively. Approximately 76% of all patients had subretinal fluid (SRF) present at baseline, and the average baseline SRF thickness (including patients with no SRF) was 120.0 μm and 129.7 μm (median, 109.5 μm and 111.0 μm) in the 0.5 mg monthly and 0.5 mg PRN groups, respectively. In these retrospective, exploratory analyses of the HARBOR study, 6 of 44 variables were found to be significant baseline predictors of visual outcomes and/or injection frequency ( Table ).
|Variables that were baseline predictors of visual outcomes at month 12|
|Age; BCVA; area of occult CNV with no classic component; total CNV leakage area; subretinal fluid presence a|
|Variables that were baseline predictors of treatment frequency in the first 12 months|
|Subretinal fluid thickness a|
|Variables that were NOT baseline predictors of visual outcomes or treatment frequency|
Current smoking status
Treatment regimen (monthly vs PRN)
Central foveal thickness
Central subfield thickness
Area of classic CNV
Percentage of classic CNV area
Evidence of fluid from CNV
Percentage of total CNV area
Area of subretinal fibrous tissue
Percentage of subretinal fibrous tissue
|Greatest linear dimension of lesion |
Percentage of occult CNV with no classic component
Area of other lesion component with no CNV
Presence of blood
Presence of classic CNV
Presence of fibrous tissue
Presence of neovascular AMD
Presence of occult CNV
Presence of photocoagulation scar
Presence of serous PED
|Total macular volume |
Total area of CNV
Total area of lesion
Macula dry b
Macula dry-2 b
Macula totally dry b
Macula totally dry-2 b
Retina dry b
Retina dry-2 b
Retinal fluid atrophy (none, extrafoveal, remote, subfoveal)
RPE rip (none, outside, b remote, b subfoveal)
Baseline Predictors for Best-Corrected Visual Acuity Change at Month 12
Significant baseline predictors for BCVA change at month 12 in patients who had both baseline and month 12 BCVA values included baseline BCVA, age, and total CNV leakage area and area of occult CNV.
Lower (ie, worse) BCVA at baseline predicted greater mean gains in BCVA at month 12 ( Figure 1 ). Patients with baseline BCVA ≤44 letters (approximate Snellen equivalent of 20/125; n = 103; gray circles) gained an average of 15.3 letters at month 12 compared with a mean gain of 8.2 letters in patients with baseline BCVA >44 letters (approximate Snellen equivalent of 20/125 or better; n = 397; black circles). Younger age at baseline also predicted greater mean gain in BCVA at month 12 ( Figure 2 ). Patients aged ≤73 years at baseline (n = 116; gray circles) gained a mean of 13.1 letters at month 12 compared with a mean gain of 8.6 letters in patients aged >73 years at baseline (n = 384; black circles).
Smaller total CNV leakage area at baseline and area of occult CNV were also predictive of greater mean gains in BCVA at month 12 ( Figure 3 ). Patients with baseline total CNV leakage area ≤5.24 DA and baseline area of occult CNV ≤5.47 DA at baseline (n = 415; Left panel) had a greater mean gain in BCVA at month 12 (+10.7 letters) than did patients with total CNV leakage area >5.24 DA and area of occult CNV >5.47 DA (n = 51; +8.9 letters; Middle panel). Patients with smaller total CNV leakage area ≤5.24 DA and area of occult CNV ≤5.47 DA at baseline also had greater mean gains than did patients with total CNV leakage area >5.24 DA and area of occult CNV ≤5.47 DA (n = 34; −3.0 letters; Right panel).
Baseline Predictors for Best-Corrected Visual Acuity Gain ≥15 Letters at Month 12
Significant baseline predictors for BCVA gain ≥15 letters at month 12 in patients who had both baseline and month 12 BCVA values included BCVA and total CNV leakage area with SRF present.
Lower (ie, worse) BCVA at baseline predicted a greater likelihood of gaining ≥15 letters at month 12 ( Figure 4 ). Among patients with baseline BCVA ≤68 letters (approximate Snellen equivalent of 20/40 or worse; n = 438), 37% gained ≥15 letters at month 12 vs 11% of patients with baseline BCVA >68 letters (approximate Snellen equivalent better than 20/40; n = 62; odds ratio [OR], 4.6; 95% confidence interval [CI] 2.1–10.5; from logistic regression model with baseline BCVA, baseline total CNV leakage area, and SRF presence at baseline as covariates).
Smaller total CNV leakage area with SRF present at baseline also predicted a greater likelihood of BCVA gain ≥15 letters at month 12 ( Figure 5 ). In patients with SRF present at baseline, 41% of those with baseline total CNV leakage area ≤4.51 DA (n = 282) gained ≥15 letters at month 12 vs 22% of patients with baseline total CNV leakage area >4.51 DA (n = 99; OR, 2.5; 95% CI 1.5–4.3; from logistic regression model with baseline BCVA, baseline total CNV leakage area, and SRF presence at baseline as covariates). In patients who did not have SRF present at baseline, total CNV leakage area at baseline was not predictive of BCVA gain ≥15 letters at month 12.